Six lines of mice were fed distinctive fixed or shifting diets while being selected for fast growth in the first two weeks after weaning. The intention was to provide alternative routes for the improvement of growth with respect to a maize diet (100% maize) where controls grew 6 g, an optimum diet (76% maize, 24% milk) where controls grew 17 g, and a ‘milk’ diet (16% maize, 84% milk) where controls grew 12 g. After 10 generations of selection the lines, hybrids and unselected controls were compared on these and intermediate diets. Realized genetic correlations between growth on optimal and suboptimal diets depended on the feeding regime in which selection was practised, and were significantly higher in lines that were selected on suboptimal diets. When the selected lines were tested on unfamiliar suboptimal diets they were hardly better than controls, but on their own diets and also on the optimum diet were about 6 g heavier. Peak growth was never much greater than on 76% maize and was made on diets containing between 58% and 82% maize. With one exception, the hybrids grew on each test diet about as well as did the best selected line.

A genetic limitation was encountered on the optimum diet. The growth there of the mice selected on either milk or maize diets was not improved upon by the mice actually selected on the optimum diet.